CardioAnnotate ECG Toolkit

1)Signal: Load Local WFDB Files (.hea + .dat/.mat) or CSV

Virtual dir: /home/pyodide/records/

Pick matching header (.hea) and signal (.dat / .mat) with the same basename, e.g. 100.hea + 100.dat.

Import from CSV: Import bundle (.npz):

2) Signal Viewer & Annotations

Channel Mode

Quick Classify R-peak:

Snap peaks to local max (±samples):

Peak detection

Run automatic detection on full signal or visible segment, assign results to a peak type.

Assign to R subtype Method Scope Replace existing (in scope)

Polarity Min distance ms Threshold (signal units) Prominence (approx.) Prom window ms

R-peaks P-peaks Q-waves S-waves T-peaks U-waves

Classify R-peak

3) Metadata, Annotations Import & Export

Import annotations:

Import WFDB annotations:

Signal: original

Import bundle (.npz):

Session (full state): Format

Name

Annotation format:

All peaks in one file (.atr) Separate files (.atr, .pwave, .twave, etc.)

4) Signal Operations

Resample to Hz

Resampling changes fs. Annotations remain at old indices until reprojected.

Filters (Butterworth, zero-phase):

High-pass Hz Low-pass Hz

Notch

Or band-pass:

– Hz

5) Stats & Log

R-peaks

P-peaks

Q-waves

S-waves

T-peaks

U-waves

6) Statistics

Signal quality (visible segment)

Computed on the currently visible window of the active channel. pSQI is QRS-band relative power; basSQI flags baseline wander; lineSQI estimates mains interference; kSQI is Pearson kurtosis (Gaussian ≈ 3).

Mains ±

pSQI (5–15 / 5–40)

—

kSQI (kurtosis)

—

basSQI (1 − 0–1 / 0–40)

—

lineSQI (mains ratio)

—

Duration

—

Avg RR (ms)

—

Avg HR (bpm)

—

Avg PR (ms)

—

Avg QT (ms)

—

Avg QRS (ms)

—

RR error detection & correction (visible segment)

Flags implausible RR intervals using absolute bounds and/or beat-to-beat rate-of-change rules, then optionally replaces flagged intervals via Gaussian process regression trained on the remaining RR intervals.

Mode GP kernel

Abs min RR (ms)

Abs max RR (ms)

Max rel change α

Compare to last accepted

Flagged RR (count)

—

Correction

—

HRV (visible segment)

Computed from R-peaks within the current zoom window. Requires ≥2 R-peaks (time-domain) and ≥4 RR intervals (frequency-domain).

Segment: —–—s

Mean NN / RR (ms)

—

SDNN (ms)

—

RMSSD (ms)

—

NN50

—

pNN50 (%)

—

Poincaré SD1 (ms)

—

Poincaré SD2 (ms)

—

SD2/SD1

—

SampEn

—

ULF power (ms²)

—

VLF power (ms²)

—

LF power (ms²)

—

HF power (ms²)

—

LF/HF ratio

—

RR Stationarity (ADF)

—